- Email: [email protected]
Radial and longitudinal variations in the Young's modulus of a Zr55Al10Ni5Cu30 bulk metallic glass rod
Materials Science and Engineering A 534 (2012) 459–464
Contents lists available at SciVerse ScienceDirect
Materials Science and Engineering A journal homepage: www.elsevier.com/locate/msea
Radial and longitudinal variations in the Young’s modulus of a Zr55 Al10 Ni5 Cu30 bulk metallic glass rod Tokujiro Yamamoto ∗ , Hisamichi Kimura, Akihisa Inoue Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan
a r t i c l e
i n f o
Article history: Received 14 October 2011 Received in revised form 24 November 2011 Accepted 28 November 2011 Available online 7 December 2011 Keywords: Bulk amorphous alloys Rapid solidification Nanoindentation
a b s t r a c t Radial and longitudinal variations in the Young’s modulus of a Zr55 Al10 Ni5 Cu30 metallic glass cylindrical rod were examined by means of nanoindentation. The metallic glass rod, 2.4 mm in diameter, was prepared using a novel mold-casting method designed to suppress crystallization resulting from the heterogeneous nucleation originating from slag covering a molten alloy. In this study, Young’s modulus was measured using a multiple partial unloading technique with a spherical indenter. In general, it is assumed that the cooling rate of a metallic glass rod depends on the radius of its cross section. However, no significant difference was found in the measured Young’s modulus. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Amorphous alloys have isotropic properties that originate from their noncrystalline random atomic arrangement. They exhibit distinctive mechanical properties, such as high strength, low Young’s modulus, and large elastic strain, compared with conventional crystalline metallic materials. However, amorphous alloys have not been widely used because they do not form bulky materials. Recently, some amorphous alloys, which are called metallic glasses, have been found exhibit glass transition. A number of molten alloys have been reported to form bulk metallic glasses by casting, because a supercooled liquid state of metallic glasses is more stable than that of conventional amorphous alloys. In particular, Zr- and Pd-based metallic glasses can form bulk specimens greater than 10 mm in diameter, even if they are cooled slowly [1–3]. The distinctive isotropic and mechanical properties of bulk metallic glasses are promising. These properties act as standard specimens for calibrating a wide range of testing equipment, in particular, for testing the mechanical properties of materials used in micro-electro-mechanical system (MEMS) devices. However, the mechanical properties of a bulk metallic glass specimen are expected to vary from place to place because the cooling rate during the casting of a specimen is not uniform. In addition, variations in properties may influence calibration if the specimen is used as a standard. To use bulk metallic glasses as standard specimens for calibrating equipment for MEMS devices, their properties
∗ Corresponding author. Tel.: +81 22 215 2159; fax: +81 22 215 2159. E-mail address: [email protected] (T. Yamamoto). 0921-5093/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2011.11.094
need to be ascertained by the equipment to confirm whether the variations in the properties of bulk metallic glasses are detectable. In this study, radial and longitudinal variations in the Young’s modulus, which is one of the mechanical properties of a Zr-based bulk metallic glass cylindrical rod, were investigated by using nanoindentation equipment.
2. Experimental procedures Ingots of Zr55 Al10 Ni5 Cu30 composition were prepared by Ar arc-melting. The ingots were crushed into pieces and subjected to modified mold-casting using a Cu horizontal mold in an Ar atmosphere. The pieces were placed in a silica nozzle and melted using high-frequency induction heating. Then, the molten alloy was injected into the Cu horizontal mold using pressurized Ar gas. The cross section of the Cu mold used in this study is shown in Fig. 1. The injected molten alloy was first stored in a reservoir. At this point, the molten alloy was covered with slag consisting of oxides and contamination. When the reservoir became full, the slag was ripped apart by the pressurized Ar gas, and the molten alloy stored in the reservoir began to flow into the horizontal mold. The molten alloy was rapidly quenched as it flowed into the horizontal mold, and a shiny cylindrical rod of 2.4 mm in diameter was formed. It is suspected that the ripped slag remained trapped at the edges of the opening between the reservoir and the horizontal mold and that crystallization by heterogeneous nucleation was suppressed. The flow rate between the bottom surface and the top surface of the molten alloy in the horizontal mold was almost the same because the tip of the rod obtained was spherical.
reservoir
2nd cycle
623 K 10.8 ks
623 K 10.8 ks
8.33 K/s
1st cycle 8.33 K/s
8.33 K/s
Cu mold
Temperature
molten alloy
orifice
silica nozzle
8.33 K/s
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
Ar gas
460
horizontal mold (φ2.4)
303 K
303 K
slag stopper
Time
In this study, the length of the cylindrical rod prepared by this method was approximately 80 mm. The rod was then sliced into 1-mm thick disks, as shown in Fig. 2. The disks were then polished using colloidal silica powder. The cross sections of the disks were examined by X-ray diffraction (XRD) using Cu K␣ radiation. The crystallization temperature Tx and the glass transition temperature Tg of one of the disks sliced at position B (see Fig. 2), 10 mm from the tip of the rod, were measured using a power-compensationtype differential scanning calorimeter (DSC) (Diamond DSC, PerkinElmer, MA, United States) at a heating rate of 0.67 K/s. The depth profiles of the distribution of several elements – Cu, C, N, and O – in the polished surface layer of another disk at position B were examined at every 1 nm in depth by means of Auger electron spectroscopy (AES) with Ar sputtering to ascertain the effect of polishing on the Young’s modulus. The intensity of Auger electrons of CuLMM at 920 eV, CKLL at 272 eV, NKLL at 381 eV, and OKLL at 510 eV was monitored. The Auger electrons of Zr, Al, Ni, Si were not monitored because of overlapping of the Auger electron energy with other elements or because they were very low in intensity for analysis. In this study, six disks sliced at positions F, G, and H (30, 35, and 40 mm from the tip of the rod, respectively) were subjected to isothermal DSC measurement or nanoindentation. The endothermic heat flow into the disks was measured, using isothermal DSC measurement, for the disks neighboring those subjected to nanoindentation. The temperature history programmed for the isothermal DSC measurement is schematically shown in Fig. 3. The specimens were rapidly heated at a rate of 8.33 K/s and then maintained at 623 K for 10.8 ks, as programmed. However, the cooling rate was slower than intended. Isothermal DSC measurements were performed twice for each specimen, with the endothermic heat flow for the second measurement being used to provide a baseline. The reduced Young’s modulus Er of the specimens was measured by means of a multiple partial unloading technique with a spherical indenter proposed by Field and Swain [4–6], using a nanoindentation measurement device (UMIS-2000, CSIRO, Lindfield, Australia). In this study, the radius of the spherical indenter used was 5 m, and the maximum load applied was 0.8 mN. The
frame compliance of 0.3 nm/mN was used to analyze the indentation results. The equipment was calibrated using fused silica as a standard specimen, whose Er and Poisson’s ratio were 69.6 GPa and 0.17, respectively. The indents made by the measurement were observed by using a scanning electron microscope (SEM).
N M L K J I H G F E D C B A
XRD isothermal DSC
nanoindentation
3. Results 3.1. Specimen characterization Fig. 4 shows the XRD pattern of the cross sections of disks and the DSC trace of one of the disks sliced at position B. The halo peaks in the XRD pattern show that the disk has a random atomic arrangement. In the DSC trace, endothermic heat was detected prior to crystallization, indicating that the disk exhibited glass transition. Thus, the disks were confirmed to be glassy. Tx and Tg are 773 K and 682 K, respectively. Fig. 5 shows the depth profiles of Cu, C, N, and O elements in the polished surface layer of another disk B. The top surface within 20 nm includes more light elements, which resulted from polishing compounds or contamination, than the deeper part of the specimen.
Intensity (a.u.)
a
20
40
60
80
100
2θ (°)
b Heat flow endo up (W/g)
10 mm
Fig. 3. Temperature history programmed for isothermal DSC measurement.
40 mm 35 mm 30 mm
Fig. 1. Cross section of a Cu horizontal mold used for casting a metallic glass rod.
500
Tx: 773 K Tg: 682 K
1 W/g 600
700
800
Temperature (K) Fig. 2. Disks sliced from the metallic glass rod. Disk B was used to measure Tx and Tg . Disks F, G, and H were subjected to XRD, isothermal DSC and nanoindentation measurements.
Fig. 4. (a) XRD pattern of disks subjected to mechanical testing, and (b) DSC trace of another disk.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
40
50
Fig. 5. Depth profiles of Cu, C, N, and O elements in the polished surface layer of one of the disks.
3.2. Isothermal DSC measurement Fig. 6 shows the endothermic heat flow and sample temperature of disks F, G, and H measured by isothermal DSC measurement. The disks subjected to isothermal DSC measurement were maintained at 623 K, which is 59 K lower than Tg . The samples were heated continuously until the temperature reached 623 K. Thus, a large endothermic heat flow to increase the sample temperature was
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
0.5
0.0
-0.5
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
0.5
0.0
-0.5
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
620 0.0
0
10
20
30 40 Time (s)
50
600 60
650
1.0
640 630
0.5 620 0.0 -0.5
700
1.5
600
610
(b')
600
650
0
630
1.5
(c)
640
0.5
700
650
0
1.0
-0.5
(b)
650
(a ')
600
Sample temperature (K)
0
Sample temperature (K)
650
0.0
Sample temperature (K)
Heat flow endo up (W/g)
(a)
-0.5
1.5
700
0.5
610 0
10
20
30 40 Time (s)
50
600 60
650
(c') 1.0
640 630
0.5 620 0.0 -0.5
Sample temperature (K)
30
Sample temperature (K)
20
Depth (nm)
Heat flow endo up (W/g)
10
Heat flow endo up (W/g)
0
detected until the temperature reached 623 K. For the first 5 s after the beginning of rapid heating to 623 K, the endothermic heat flow for both the first and second measurement reached more than 1.0 W/g. After 5 s, the endothermic heat flow for the second measurement decreased gradually and monotonically. In contrast, the heat flow for the first measurement reached its minimum at approximately 14 s. This difference between the first and second trace of the endothermic heat flow was a result of structural relaxation. The disks relaxed considerably after being maintained at 623 K for 10.8 ks during the first isothermal DSC measurement. Exothermic heat due to structural relaxation at 623 K was barely detected in the second isothermal DSC measurement. Therefore, the subtraction of the second heat flow from the first heat flow represents the heat generated by structural relaxation of the disks. For both the first and second isothermal DSC measurements, the sample temperature became stable 12 s after the beginning of rapid heating, and the heat flow into the disks became stable after 7.2 ks. Thus, the subtracted endothermic heat between the first and second measurements was integrated from 12 s to 7.2 ks in order to compare the structural relaxation in disks F, G, and H. The integrated endothermic heat of disks F, G, and H were 12.1, 18.5, and 7.1 J/g, respectively. The integrated endothermic heat represents the progress of structural relaxation in the disks, although the integrated heat does not contain the heat generated by structural relaxation before 12 s. The larger the integrated heat becomes, the
610 0
10
20
30 40 Time (s)
50
Sample temperature (K)
Intensity (a.u.)
Cu C N O
461
600 60
Heat flow endo up (1st) Heat flow endo up (2nd) Heat flow endo up (subtracted) Sample temperature (1st) Sample temperature (2nd) Fig. 6. Isothermal DSC traces and its magnifications of metallic glass disks (a, a ) F, (b, b ) G, and (c, c ) H used for the first and second measurements as a function of elapsed time after the beginning of rapid heating. Traces of the sample temperature during measurements are also shown, but they overlap one another.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
Reduced modulus, Er (GPa)
462
the metallic glass disks is presumed to be more structurally relaxed than the outer shell because bulk metallic glasses are cooled from the outside during quenching; thus, the cooling rate in the inner core is slower than that in the outer shell. However, no clear difference in Er was observed between the inner core and the outside shell of any of the disks.
100 80 60 40
4. Discussion
20 0 0
50
100
150
Depth, hc (nm) Fig. 7. Er of the fused silica measured at 5 points.
more the specimen relaxes structurally. In this study, the cooling rate of the molten alloy flowing in the horizontal mold is assumed to vary monotonically in the longitudinal direction of the rod. However, the integrated heat of the disks did not change monotonically as a function of the distance from the tip. Thus, the progress of structural relaxation in the metallic glass rod was not detectable by isothermal DSC measurements. 3.3. Multiple partial unloading with a spherical indenter Fig. 7 shows five sets of Er of the fused silica as a function of indentation depth, hc , where hc is the penetration depth below the contact circle by the indenter and sample [4–6]. Except for the surface layer within 20 nm, the Er measured by the multiple partial unloading technique is almost the same as 69.6 GPa, which is the Er of the fused silica because the experimental result was successfully calibrated by using the parameter for the shape of the tip of the spherical indenter [4–6]. Fig. 8 shows the marks made by the spherical indenter under loads of 1.0 mN, 5.0 mN, and 10.0 mN using another metallic glass disk. When the load was greater than 5.0 mN, the disk deformed plastically, and the surface was piled up around the indent. Thus, a load of 0.8 mN was chosen for nanoindentation in this study to prevent the plastic deformation of the specimen as much possible. Fig. 9 shows five sets of calculated Er values as a function of hc for the F, G, and H metallic glass disks. Nanoindentation tests were performed in the area approximately 0.2–0.8 mm away from the center of each disk. In accord with the presence of the surface layer containing more light elements, Er gradually increased with increasing hc . After hc reaches 20 nm, Er ranges from 80 to 110 GPa. The Er of the disks is approximately 90–100 GPa for hc deeper than 80 nm, and close to the Young’s modulus of a Zr55 Al10 Ni5 Cu30 bulk metallic glass reported previously [7,8]. However, the nanoindentation test could not detect the difference of Er among disks F, G, and H, as well as isothermal DSC measurement. The inner core of
The subtracted endothermic heat flow traces indicate that most of the exothermic heat caused by structural relaxation at 623 K, which is 59 K lower than Tg of Zr55 Al10 Ni5 Cu30 , is clearly detected within 1 ks, as shown in Figs. 6(a)–(c). However, Haruyama et al. [9] showed by density measurement that more than 6 ks is necessary for a Zr50 Cu40 Al10 bulk metallic glass to become structurally relaxed at 653 K, which is 44 K lower temperature than Tg of the metallic glass they used. The time required for structural relaxation in this study seems much shorter than that for Zr50 Cu40 Al10 bulk metallic glass. One of the reasons for the shorter relaxation time in this study is the difficulty in detecting a small heat flow by using DSC. The sensitivity of the DSC to detect heat flow caused by structural relaxation is assumed to be poorer than the evaluation of structural relaxation made using density measurements. Another reason is the slow cooling rate during casting when a horizontal mold is used. In the literature [9], a conventional vertical mold was probably used to cast the Zr50 Cu40 Al10 bulk metallic glass. Fig. 10 shows schematic illustrations of the vitrification process during casting using the conventional vertical mold and horizontal mold, as used in this study. When using a vertical mold for casting, a molten alloy is poured into the mold in one shot; thus, the molten alloy is quenched immediately (Fig. 10(a)). When the temperature of the molten alloy reaches Tg , the alloy starts vitrification, starting from the surface in contact with the mold. On the other hand, a molten alloy cast in the horizontal mold is quenched slowly. When the molten alloy flows from the reservoir into the horizontal mold by ripping the slag, it is rapidly cooled from its surface. However, the inside encounters difficulty in cooling because of the continuous supply of the hot molten alloy from the reservoir (Fig. 10(b)(i)). The tip of the flowing molten alloy is cooled slowly because it is not in contact with the mold. When the temperature of the tip of the flowing molten alloy approaches Tg and its viscosity increases, the molten alloy with high viscosity at the tip is pushed aside by the following molten alloy with low viscosity. Then, the tip of the flowing molten alloy continues to proceed in the horizontal mold (Fig. 10(b)(ii)). Finally, the tip of the molten alloy stops proceeding when it is cooled to Tg . Furthermore, the molten alloy stops flowing and starts vitrifying (Fig. 10(b)(iii)). As a result, the cooling rate of a metallic glass rod cast in a horizontal mold is considerably slower than that of bulk metallic glasses cast in a conventional vertical mold.
Fig. 8. Composite SEM image of the marks made by the spherical indenter under loads of 1.0, 5.0, and 10.0 mN.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
463
φ2.4 mm metallic glass disk
Center 0.2 mm 0.4 mm 0.6 mm 0.8 mm
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
20
40
60
80
100
0
20
40
60
80
100
0
20
40
60
80
100
the center
50
the center
50
0.4 mm from
50
the center
100
0.6 mm from
100
the center
100
0.2 mm from
Disk H 150
0
Reduced modulus, Er (GPa)
Disk G 150
0.8 mm from
Disk F 150
Depth, hc (nm) Fig. 9. Er of the metallic glass disks F, G, and H measured at 5 points in areas that are isolated every 0.2 mm from the center.
464
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
structural relaxation of the metallic glass rod was ascertained by isothermal DSC measurement, and Er of the metallic glass rod was measured using a multiple partial unloading technique with a spherical indenter. Er increased gradually with increasing hc and approached approximately 90–100 GPa, independent of the location in the metallic glass rod. It was difficult to detect the progress of structural relaxation in the metallic glass rod by isothermal DSC measurement or by nanoindentation. The cooling history of the bulk metallic glass rod cast in a horizontal mold is much slower than that in a conventional vertical mold. The variation in Young’s modulus of the rod was negligible because the rod was structurally relaxed by the slower cooling rate.
glass
molten alloy with high viscosity
molten alloy
a
mold
Acknowledgments
b (i)
(ii)
(iii)
molten alloy
glass
mold
molten alloy with high viscosity Fig. 10. Schematic illustration of vitrification during casting in (a) a conventional vertical mold and (b) a horizontal mold.
5. Conclusions A Zr55 Al10 Ni5 Cu30 metallic glass rod was prepared by casting using a horizontal mold. The exothermic heat generated by
The authors wish to thank Dr. Motohiro Suganuma of the Aichi Industrial Technology Institute for his advice on nanoindentation. The authors also thank technical assistants Mr. Kanomata and Mr. Murakami at IMR, Tohoku University. This work was supported in part by a Grant-in-Aid for Young Scientists (B) 21760548 and Challenging Exploratory Research 23656421 from the Japan Society for the Promotion of Science (JSPS) and by the Amada Foundation for Metal Work Technology. References [1] [2] [3] [4] [5] [6] [7] [8] [9]
A. Peker, W.L. Johnson, Appl. Phys. Lett. 63 (1993) 2342. A. Inoue, N. Nishiyama, T. Matsuda, Mater. Trans. JIM 37 (1996) 181. A. Inoue, T. Zhang, Mater. Trans. JIM 37 (1996) 185. J.S. Field, M.V. Swain, J. Mater. Res. 8 (1993) 297. J.S. Field, M.V. Swain, J. Mater. Res. 10 (1995) 101. M. Suganuma, M.V. Swain, J. Mater. Res. 19 (2004) 3490. M.L. Vaillant, V. Keryvin, T. Rouxel, Y. Kawamura, Scripta Mater. 47 (2002) 19. Y. Yokoyama, K. Fukaura, A. Inoue, Mater. Trans. 45 (2004) 1672. O. Haruyama, Y. Yokoyama, A. Inoue, Mater. Trans. 48 (2007) 1708.
Contents lists available at SciVerse ScienceDirect
Materials Science and Engineering A journal homepage: www.elsevier.com/locate/msea
Radial and longitudinal variations in the Young’s modulus of a Zr55 Al10 Ni5 Cu30 bulk metallic glass rod Tokujiro Yamamoto ∗ , Hisamichi Kimura, Akihisa Inoue Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan
a r t i c l e
i n f o
Article history: Received 14 October 2011 Received in revised form 24 November 2011 Accepted 28 November 2011 Available online 7 December 2011 Keywords: Bulk amorphous alloys Rapid solidification Nanoindentation
a b s t r a c t Radial and longitudinal variations in the Young’s modulus of a Zr55 Al10 Ni5 Cu30 metallic glass cylindrical rod were examined by means of nanoindentation. The metallic glass rod, 2.4 mm in diameter, was prepared using a novel mold-casting method designed to suppress crystallization resulting from the heterogeneous nucleation originating from slag covering a molten alloy. In this study, Young’s modulus was measured using a multiple partial unloading technique with a spherical indenter. In general, it is assumed that the cooling rate of a metallic glass rod depends on the radius of its cross section. However, no significant difference was found in the measured Young’s modulus. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Amorphous alloys have isotropic properties that originate from their noncrystalline random atomic arrangement. They exhibit distinctive mechanical properties, such as high strength, low Young’s modulus, and large elastic strain, compared with conventional crystalline metallic materials. However, amorphous alloys have not been widely used because they do not form bulky materials. Recently, some amorphous alloys, which are called metallic glasses, have been found exhibit glass transition. A number of molten alloys have been reported to form bulk metallic glasses by casting, because a supercooled liquid state of metallic glasses is more stable than that of conventional amorphous alloys. In particular, Zr- and Pd-based metallic glasses can form bulk specimens greater than 10 mm in diameter, even if they are cooled slowly [1–3]. The distinctive isotropic and mechanical properties of bulk metallic glasses are promising. These properties act as standard specimens for calibrating a wide range of testing equipment, in particular, for testing the mechanical properties of materials used in micro-electro-mechanical system (MEMS) devices. However, the mechanical properties of a bulk metallic glass specimen are expected to vary from place to place because the cooling rate during the casting of a specimen is not uniform. In addition, variations in properties may influence calibration if the specimen is used as a standard. To use bulk metallic glasses as standard specimens for calibrating equipment for MEMS devices, their properties
∗ Corresponding author. Tel.: +81 22 215 2159; fax: +81 22 215 2159. E-mail address: [email protected] (T. Yamamoto). 0921-5093/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2011.11.094
need to be ascertained by the equipment to confirm whether the variations in the properties of bulk metallic glasses are detectable. In this study, radial and longitudinal variations in the Young’s modulus, which is one of the mechanical properties of a Zr-based bulk metallic glass cylindrical rod, were investigated by using nanoindentation equipment.
2. Experimental procedures Ingots of Zr55 Al10 Ni5 Cu30 composition were prepared by Ar arc-melting. The ingots were crushed into pieces and subjected to modified mold-casting using a Cu horizontal mold in an Ar atmosphere. The pieces were placed in a silica nozzle and melted using high-frequency induction heating. Then, the molten alloy was injected into the Cu horizontal mold using pressurized Ar gas. The cross section of the Cu mold used in this study is shown in Fig. 1. The injected molten alloy was first stored in a reservoir. At this point, the molten alloy was covered with slag consisting of oxides and contamination. When the reservoir became full, the slag was ripped apart by the pressurized Ar gas, and the molten alloy stored in the reservoir began to flow into the horizontal mold. The molten alloy was rapidly quenched as it flowed into the horizontal mold, and a shiny cylindrical rod of 2.4 mm in diameter was formed. It is suspected that the ripped slag remained trapped at the edges of the opening between the reservoir and the horizontal mold and that crystallization by heterogeneous nucleation was suppressed. The flow rate between the bottom surface and the top surface of the molten alloy in the horizontal mold was almost the same because the tip of the rod obtained was spherical.
reservoir
2nd cycle
623 K 10.8 ks
623 K 10.8 ks
8.33 K/s
1st cycle 8.33 K/s
8.33 K/s
Cu mold
Temperature
molten alloy
orifice
silica nozzle
8.33 K/s
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
Ar gas
460
horizontal mold (φ2.4)
303 K
303 K
slag stopper
Time
In this study, the length of the cylindrical rod prepared by this method was approximately 80 mm. The rod was then sliced into 1-mm thick disks, as shown in Fig. 2. The disks were then polished using colloidal silica powder. The cross sections of the disks were examined by X-ray diffraction (XRD) using Cu K␣ radiation. The crystallization temperature Tx and the glass transition temperature Tg of one of the disks sliced at position B (see Fig. 2), 10 mm from the tip of the rod, were measured using a power-compensationtype differential scanning calorimeter (DSC) (Diamond DSC, PerkinElmer, MA, United States) at a heating rate of 0.67 K/s. The depth profiles of the distribution of several elements – Cu, C, N, and O – in the polished surface layer of another disk at position B were examined at every 1 nm in depth by means of Auger electron spectroscopy (AES) with Ar sputtering to ascertain the effect of polishing on the Young’s modulus. The intensity of Auger electrons of CuLMM at 920 eV, CKLL at 272 eV, NKLL at 381 eV, and OKLL at 510 eV was monitored. The Auger electrons of Zr, Al, Ni, Si were not monitored because of overlapping of the Auger electron energy with other elements or because they were very low in intensity for analysis. In this study, six disks sliced at positions F, G, and H (30, 35, and 40 mm from the tip of the rod, respectively) were subjected to isothermal DSC measurement or nanoindentation. The endothermic heat flow into the disks was measured, using isothermal DSC measurement, for the disks neighboring those subjected to nanoindentation. The temperature history programmed for the isothermal DSC measurement is schematically shown in Fig. 3. The specimens were rapidly heated at a rate of 8.33 K/s and then maintained at 623 K for 10.8 ks, as programmed. However, the cooling rate was slower than intended. Isothermal DSC measurements were performed twice for each specimen, with the endothermic heat flow for the second measurement being used to provide a baseline. The reduced Young’s modulus Er of the specimens was measured by means of a multiple partial unloading technique with a spherical indenter proposed by Field and Swain [4–6], using a nanoindentation measurement device (UMIS-2000, CSIRO, Lindfield, Australia). In this study, the radius of the spherical indenter used was 5 m, and the maximum load applied was 0.8 mN. The
frame compliance of 0.3 nm/mN was used to analyze the indentation results. The equipment was calibrated using fused silica as a standard specimen, whose Er and Poisson’s ratio were 69.6 GPa and 0.17, respectively. The indents made by the measurement were observed by using a scanning electron microscope (SEM).
N M L K J I H G F E D C B A
XRD isothermal DSC
nanoindentation
3. Results 3.1. Specimen characterization Fig. 4 shows the XRD pattern of the cross sections of disks and the DSC trace of one of the disks sliced at position B. The halo peaks in the XRD pattern show that the disk has a random atomic arrangement. In the DSC trace, endothermic heat was detected prior to crystallization, indicating that the disk exhibited glass transition. Thus, the disks were confirmed to be glassy. Tx and Tg are 773 K and 682 K, respectively. Fig. 5 shows the depth profiles of Cu, C, N, and O elements in the polished surface layer of another disk B. The top surface within 20 nm includes more light elements, which resulted from polishing compounds or contamination, than the deeper part of the specimen.
Intensity (a.u.)
a
20
40
60
80
100
2θ (°)
b Heat flow endo up (W/g)
10 mm
Fig. 3. Temperature history programmed for isothermal DSC measurement.
40 mm 35 mm 30 mm
Fig. 1. Cross section of a Cu horizontal mold used for casting a metallic glass rod.
500
Tx: 773 K Tg: 682 K
1 W/g 600
700
800
Temperature (K) Fig. 2. Disks sliced from the metallic glass rod. Disk B was used to measure Tx and Tg . Disks F, G, and H were subjected to XRD, isothermal DSC and nanoindentation measurements.
Fig. 4. (a) XRD pattern of disks subjected to mechanical testing, and (b) DSC trace of another disk.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
40
50
Fig. 5. Depth profiles of Cu, C, N, and O elements in the polished surface layer of one of the disks.
3.2. Isothermal DSC measurement Fig. 6 shows the endothermic heat flow and sample temperature of disks F, G, and H measured by isothermal DSC measurement. The disks subjected to isothermal DSC measurement were maintained at 623 K, which is 59 K lower than Tg . The samples were heated continuously until the temperature reached 623 K. Thus, a large endothermic heat flow to increase the sample temperature was
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
0.5
0.0
-0.5
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
0.5
0.0
-0.5
2
4 6 Time (ks)
8
10
Heat flow endo up (W/g)
620 0.0
0
10
20
30 40 Time (s)
50
600 60
650
1.0
640 630
0.5 620 0.0 -0.5
700
1.5
600
610
(b')
600
650
0
630
1.5
(c)
640
0.5
700
650
0
1.0
-0.5
(b)
650
(a ')
600
Sample temperature (K)
0
Sample temperature (K)
650
0.0
Sample temperature (K)
Heat flow endo up (W/g)
(a)
-0.5
1.5
700
0.5
610 0
10
20
30 40 Time (s)
50
600 60
650
(c') 1.0
640 630
0.5 620 0.0 -0.5
Sample temperature (K)
30
Sample temperature (K)
20
Depth (nm)
Heat flow endo up (W/g)
10
Heat flow endo up (W/g)
0
detected until the temperature reached 623 K. For the first 5 s after the beginning of rapid heating to 623 K, the endothermic heat flow for both the first and second measurement reached more than 1.0 W/g. After 5 s, the endothermic heat flow for the second measurement decreased gradually and monotonically. In contrast, the heat flow for the first measurement reached its minimum at approximately 14 s. This difference between the first and second trace of the endothermic heat flow was a result of structural relaxation. The disks relaxed considerably after being maintained at 623 K for 10.8 ks during the first isothermal DSC measurement. Exothermic heat due to structural relaxation at 623 K was barely detected in the second isothermal DSC measurement. Therefore, the subtraction of the second heat flow from the first heat flow represents the heat generated by structural relaxation of the disks. For both the first and second isothermal DSC measurements, the sample temperature became stable 12 s after the beginning of rapid heating, and the heat flow into the disks became stable after 7.2 ks. Thus, the subtracted endothermic heat between the first and second measurements was integrated from 12 s to 7.2 ks in order to compare the structural relaxation in disks F, G, and H. The integrated endothermic heat of disks F, G, and H were 12.1, 18.5, and 7.1 J/g, respectively. The integrated endothermic heat represents the progress of structural relaxation in the disks, although the integrated heat does not contain the heat generated by structural relaxation before 12 s. The larger the integrated heat becomes, the
610 0
10
20
30 40 Time (s)
50
Sample temperature (K)
Intensity (a.u.)
Cu C N O
461
600 60
Heat flow endo up (1st) Heat flow endo up (2nd) Heat flow endo up (subtracted) Sample temperature (1st) Sample temperature (2nd) Fig. 6. Isothermal DSC traces and its magnifications of metallic glass disks (a, a ) F, (b, b ) G, and (c, c ) H used for the first and second measurements as a function of elapsed time after the beginning of rapid heating. Traces of the sample temperature during measurements are also shown, but they overlap one another.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
Reduced modulus, Er (GPa)
462
the metallic glass disks is presumed to be more structurally relaxed than the outer shell because bulk metallic glasses are cooled from the outside during quenching; thus, the cooling rate in the inner core is slower than that in the outer shell. However, no clear difference in Er was observed between the inner core and the outside shell of any of the disks.
100 80 60 40
4. Discussion
20 0 0
50
100
150
Depth, hc (nm) Fig. 7. Er of the fused silica measured at 5 points.
more the specimen relaxes structurally. In this study, the cooling rate of the molten alloy flowing in the horizontal mold is assumed to vary monotonically in the longitudinal direction of the rod. However, the integrated heat of the disks did not change monotonically as a function of the distance from the tip. Thus, the progress of structural relaxation in the metallic glass rod was not detectable by isothermal DSC measurements. 3.3. Multiple partial unloading with a spherical indenter Fig. 7 shows five sets of Er of the fused silica as a function of indentation depth, hc , where hc is the penetration depth below the contact circle by the indenter and sample [4–6]. Except for the surface layer within 20 nm, the Er measured by the multiple partial unloading technique is almost the same as 69.6 GPa, which is the Er of the fused silica because the experimental result was successfully calibrated by using the parameter for the shape of the tip of the spherical indenter [4–6]. Fig. 8 shows the marks made by the spherical indenter under loads of 1.0 mN, 5.0 mN, and 10.0 mN using another metallic glass disk. When the load was greater than 5.0 mN, the disk deformed plastically, and the surface was piled up around the indent. Thus, a load of 0.8 mN was chosen for nanoindentation in this study to prevent the plastic deformation of the specimen as much possible. Fig. 9 shows five sets of calculated Er values as a function of hc for the F, G, and H metallic glass disks. Nanoindentation tests were performed in the area approximately 0.2–0.8 mm away from the center of each disk. In accord with the presence of the surface layer containing more light elements, Er gradually increased with increasing hc . After hc reaches 20 nm, Er ranges from 80 to 110 GPa. The Er of the disks is approximately 90–100 GPa for hc deeper than 80 nm, and close to the Young’s modulus of a Zr55 Al10 Ni5 Cu30 bulk metallic glass reported previously [7,8]. However, the nanoindentation test could not detect the difference of Er among disks F, G, and H, as well as isothermal DSC measurement. The inner core of
The subtracted endothermic heat flow traces indicate that most of the exothermic heat caused by structural relaxation at 623 K, which is 59 K lower than Tg of Zr55 Al10 Ni5 Cu30 , is clearly detected within 1 ks, as shown in Figs. 6(a)–(c). However, Haruyama et al. [9] showed by density measurement that more than 6 ks is necessary for a Zr50 Cu40 Al10 bulk metallic glass to become structurally relaxed at 653 K, which is 44 K lower temperature than Tg of the metallic glass they used. The time required for structural relaxation in this study seems much shorter than that for Zr50 Cu40 Al10 bulk metallic glass. One of the reasons for the shorter relaxation time in this study is the difficulty in detecting a small heat flow by using DSC. The sensitivity of the DSC to detect heat flow caused by structural relaxation is assumed to be poorer than the evaluation of structural relaxation made using density measurements. Another reason is the slow cooling rate during casting when a horizontal mold is used. In the literature [9], a conventional vertical mold was probably used to cast the Zr50 Cu40 Al10 bulk metallic glass. Fig. 10 shows schematic illustrations of the vitrification process during casting using the conventional vertical mold and horizontal mold, as used in this study. When using a vertical mold for casting, a molten alloy is poured into the mold in one shot; thus, the molten alloy is quenched immediately (Fig. 10(a)). When the temperature of the molten alloy reaches Tg , the alloy starts vitrification, starting from the surface in contact with the mold. On the other hand, a molten alloy cast in the horizontal mold is quenched slowly. When the molten alloy flows from the reservoir into the horizontal mold by ripping the slag, it is rapidly cooled from its surface. However, the inside encounters difficulty in cooling because of the continuous supply of the hot molten alloy from the reservoir (Fig. 10(b)(i)). The tip of the flowing molten alloy is cooled slowly because it is not in contact with the mold. When the temperature of the tip of the flowing molten alloy approaches Tg and its viscosity increases, the molten alloy with high viscosity at the tip is pushed aside by the following molten alloy with low viscosity. Then, the tip of the flowing molten alloy continues to proceed in the horizontal mold (Fig. 10(b)(ii)). Finally, the tip of the molten alloy stops proceeding when it is cooled to Tg . Furthermore, the molten alloy stops flowing and starts vitrifying (Fig. 10(b)(iii)). As a result, the cooling rate of a metallic glass rod cast in a horizontal mold is considerably slower than that of bulk metallic glasses cast in a conventional vertical mold.
Fig. 8. Composite SEM image of the marks made by the spherical indenter under loads of 1.0, 5.0, and 10.0 mN.
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
463
φ2.4 mm metallic glass disk
Center 0.2 mm 0.4 mm 0.6 mm 0.8 mm
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
150
150
150
100
100
100
50
50
50
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
0
20
40
60
80
100
0
20
40
60
80
100
0
20
40
60
80
100
0
20
40
60
80
100
the center
50
the center
50
0.4 mm from
50
the center
100
0.6 mm from
100
the center
100
0.2 mm from
Disk H 150
0
Reduced modulus, Er (GPa)
Disk G 150
0.8 mm from
Disk F 150
Depth, hc (nm) Fig. 9. Er of the metallic glass disks F, G, and H measured at 5 points in areas that are isolated every 0.2 mm from the center.
464
T. Yamamoto et al. / Materials Science and Engineering A 534 (2012) 459–464
structural relaxation of the metallic glass rod was ascertained by isothermal DSC measurement, and Er of the metallic glass rod was measured using a multiple partial unloading technique with a spherical indenter. Er increased gradually with increasing hc and approached approximately 90–100 GPa, independent of the location in the metallic glass rod. It was difficult to detect the progress of structural relaxation in the metallic glass rod by isothermal DSC measurement or by nanoindentation. The cooling history of the bulk metallic glass rod cast in a horizontal mold is much slower than that in a conventional vertical mold. The variation in Young’s modulus of the rod was negligible because the rod was structurally relaxed by the slower cooling rate.
glass
molten alloy with high viscosity
molten alloy
a
mold
Acknowledgments
b (i)
(ii)
(iii)
molten alloy
glass
mold
molten alloy with high viscosity Fig. 10. Schematic illustration of vitrification during casting in (a) a conventional vertical mold and (b) a horizontal mold.
5. Conclusions A Zr55 Al10 Ni5 Cu30 metallic glass rod was prepared by casting using a horizontal mold. The exothermic heat generated by
The authors wish to thank Dr. Motohiro Suganuma of the Aichi Industrial Technology Institute for his advice on nanoindentation. The authors also thank technical assistants Mr. Kanomata and Mr. Murakami at IMR, Tohoku University. This work was supported in part by a Grant-in-Aid for Young Scientists (B) 21760548 and Challenging Exploratory Research 23656421 from the Japan Society for the Promotion of Science (JSPS) and by the Amada Foundation for Metal Work Technology. References [1] [2] [3] [4] [5] [6] [7] [8] [9]
A. Peker, W.L. Johnson, Appl. Phys. Lett. 63 (1993) 2342. A. Inoue, N. Nishiyama, T. Matsuda, Mater. Trans. JIM 37 (1996) 181. A. Inoue, T. Zhang, Mater. Trans. JIM 37 (1996) 185. J.S. Field, M.V. Swain, J. Mater. Res. 8 (1993) 297. J.S. Field, M.V. Swain, J. Mater. Res. 10 (1995) 101. M. Suganuma, M.V. Swain, J. Mater. Res. 19 (2004) 3490. M.L. Vaillant, V. Keryvin, T. Rouxel, Y. Kawamura, Scripta Mater. 47 (2002) 19. Y. Yokoyama, K. Fukaura, A. Inoue, Mater. Trans. 45 (2004) 1672. O. Haruyama, Y. Yokoyama, A. Inoue, Mater. Trans. 48 (2007) 1708.